Buying Less, but Shopping More: The Use of Non- Market Labor During a Crisis David McKenzie, World Bank Ernesto Schargrodsky, Di Tella.

Buying Less, but Shopping More: The Use of Non-Market Labor During a Crisis

David McKenzie, World Bank

Ernesto Schargrodsky, Di Tella

“Women, the poor, children, the unemployed, etc. would be more willing to spend their time in a queue or otherwise ferreting out rationed goods than would high-earning males” (Becker 1965, A Theory of the Allocation of Time).

Adjustment to Aggregate Shocks Reduced set of risk-coping mechanisms

Formal and informal credit dries upRising inflation erodes value of savings Price of assets fall.group-based informal insurance can’t protect against

common shocksUnemployment limits use of labor market

=> Total expenditure falls by as much as income

Alternative Coping Strategies

Reallocate expenditure, spending more on basic foods

Take actions to change how much food a given amount of expenditure can buySpend more time searchingShop at a wider variety of storesBuy lower quality products

Non-Market Use of Labor When income falls, consumers can substitute

goods for time in the production of consumption by increasing the time devoted to shopping and other home production activities.

It has proven difficult, however, to test this implication.

Standard expenditure surveys generally provide little information on shopping consumer behavior.

When expenditure surveys including detailed shopping data exist, they usually have a cross-sectional structure with no exogenous source of income variation.

We use high-frequency (10-day) expenditure data from LatinPanel, a market research firm, to analyze shopping behavior during the 2002 Argentine crisis.

The crisis provides a large exogenous shocks to households’ income which allows to identify the causal effect of income on shopping activity.

Non-Market Use of Labor

Preview of results

Consumers are found to shop more days a week at a wider variety of channels during the crisis.

The fall in income explains much of the increase in shopping frequency.

Increased shopping frequency results in lower prices and more priced goods, mitigating up to 40% of the fall in expenditure.

Outline

Macroeconomic Overview Data Basic Facts – changes in expenditure, shopping

frequency, stores visited and quality Explaining the increase in shopping frequency Prevalence as a mitigation mechanism Conclusions

The Macro Context Recession 1998(2)-2001 December 3, 2001: corralito – partial freeze on

deposits January 6, 2002: govt. votes to end 11 years of

convertibility and float the peso Real GDP fell 10.9% in 2002, private real

consumption 14.4% Unemployment increases to 21.5% Poverty increases to 37.7%

11

.52

2.5

33

.5P

eso

s p

er

U.S

. D

olla

r

2000 2001 2002 2003 2004Year and Month

Argentine Peso - U.S. Dollar Exchange Rate2000-2003

Exchange Rate

-50

51

01

5In

flatio

n (

%)

2000 2001 2002 2003 2004Year and Month

CPI Inflation Food Inflation

LatinPanelfood inflation

Monthly Inflation Rates

Measure of Liquidity Effect of Corralito

Discount rate offered by exchange houses for exchanging money inside the corralito for cash0

51

01

5P

rem

ium

(%

)

2001 2002 2003Year and Month

Corralito Liquidity Premium

Price Dispersion Average of coefficients of variation for individual

products weighted by 2000 expenditure shares

.4.4

5.5

.55

We

igh

ted

C.V

.

2000 2001 2002 2003Year and Month

Price Dispersion

LatinPanel Data

Market research firm rotating panel of 3000 households nationwide for 2000-02

Purchase diaries kept of expenditure on food, cleaning and beauty products

Record day of purchase, amount, units, quality, channel for each 10-day period

Two main quality levels for each product: premium and priced goods

Products

Food, cleaning and beauty products with brands Fresh fruit, bread, fresh meat, and meals out not

included Mean expenditure shares: Food 76%, cleaning

13%, beauty 11% Calculate that LatinPanel food is about 50% of

total food basket, total LP expenditure is around 15% of monthly income

Mixture of necessity and luxury items

Ten Channels

Hypermarkets Supermarkets Discounts Wholesalers Drugstores

Kioscos (candy store) Almacenes (small grocery

store) Autoservicios (self-

service) Trueque (Barter Clubs) Other channels (Ferias

etc)

Pseudo-panel structure

For confidentiality restrictions LP does not provide data at the individual level, but instead aggregated into pseudo-households

Pseudos contain all households with the same demographic and socioeconomic characteristics: Region *Socioeconomic Class * household size * housewife’s age * youngest child age

In practice just over 400 pseudos Can also use this structure to match with labor

force survey

Income

Labor-force survey (EPH) provides income for matched pseudo-households: available for months of April and September each year only

Monthly average wage from Social Security system to extrapolate change in average income between EPH months.

Buying Less

Real LatinPanel total expenditure fell 10.6% between 2001 and 2002 (after 2.5% fall

between 2000 and 2001)

Quantity falls for all food (apart from yerba mate and pasta) and for all cleaning and beauty products

Expenditure on premium products fell 17.6% Expenditure on priced goods rose 2%

Shopping more days3

3.5

44.

55

5.5

Sho

ppin

g D

ays

pe

r 10

day

pe

riod

2001 2001.5 2002 2002.5 2003Year

All goods Premium Goods

Nonpremium Goods

Mean days each household spentshopping per 10 day period 2001-2002

Shopping in more places1.

61.

82

2.2

2.4

2.6

Num

ber

of C

han

nels

pe

r 1

0 d

ay p

erio

d

2000 2001 2002 2003Year

All goods Premium Goods

Priced Brand Goods

Mean number of channels shoppedper household each 10 day period

Measuring shopping frequency

Channel-days: sum of the number of days spent shopping at each of the ten channels in the 10-day period

Average pseudo-household spent 6.28 channel-days per 10-day period in 2001 and 6.71 in 2002 (7% increase)

TABLE 3: CHANGES IN SHOPPING FREQUENCY 2001-2002

AllHouseholds Lowest 2nd 3rd Highest

Channel-Days Shopped at per 10-day period 2001 6.28 6.04 6.56 6.14 5.972002 6.71** 6.68** 7.07** 6.64** 6.66**

Channel-Days Shopped per Real Peso Spent 2001 0.24 0.29 0.26 0.21 0.162002 0.28** 0.35** 0.31** 0.26** 0.20**

Households by Income Quartile

Empirical Questions

1. Do people shop more as a response to falling incomes?

2. Or is shopping more just another cost of the crisis, reflecting increases in liquidity constraints, inflation, or price dispersion?

3. If this is a mitigation mechanism, how important is it?

The Income Effect (Cross-section)

On the one hand, falling income lowers the opportunity cost of time, increasing shopping.

On the other hand, falling income reduces consumption and the gains from shopping.

Cross-sectional analysis: look at relationship between income and shopping frequency in 2001, prior to the crisis. Non-parametrics: Freq = f(Labor Income) Semi-parametrics: control also for quantities

4.5

55

.56

6.5

Da

y C

ha

nn

els

Sh

op

pe

d

3 4 5 6 7 8Real Log monthly income

Nonparametrics: Shopping Frequency against Income

22

.53

3.5

4D

ay C

ha

nn

els

Sh

op

pe

d

3 4 5 6 7 8Real Log monthly income

Semiparametrics: Shopping Frequency against Income controlling for quantity purchased

The Income Effect Over Time

Look at changes in income and shopping frequency for pseudos over time

Control for pseudo-household fixed effects to capture time-invariant determinants of shopping frequency

37

1,,,,, log

jthhhjjthtthth qXZincomeFreqShopping

TABLE 4: DETERMINANTS OF SHOPPING FREQUENCY 2001-02Dependent Variable:Shopping Frequency (Channel-days shopped at per 10 day period)

(3) (4)Fitted EPH log real labor income -0.730 -0.586

(12.53)** (9.18)**Corralito premium 0.004

(1.14)Food CPI inflation 0.043

(9.52)**Aggregate price dispersion across channels -1.099

(2.55)*

Product quantity effects yes yesPseudo-household fixed effects yes yes

Other crisis effects Corralito: restricts cash on hand, reducing

liquidity Expect this to cause people to shop more frequently

buying fewer items each time Inflation erodes nominal balances, causing

consumers to lower their reservation prices and increase shopping.

Inflation generally accompanied by Price Dispersion Consumers will want to hold a lower stock of

knowledge about prices May need to increase or decrease search to hold this

smaller stock of information

Aggregate time trends

Add time effects Allow effect of inflation, price dispersion,

and corralito to differ by householdConstruct pseudo-household CPI and price

dispersion Interact Corralito with credit card ownership

TABLE 5: ROBUSTNESS OF SHOPPING FREQUENCY REGRESSIONSDependent Variables: Channel-days shopped at per 10 day period (Columns 1-6), Channel-days per real peso per 10 day period (Column 7)

(4) (5) (6) (7)Fitted EPH log real labor income -0.237 -0.238 -0.558 -0.030

(2.83)** (2.84)** (8.67)** (9.46)**Corralito premium 0.002 -0.001

(0.56) (6.48)**Food CPI inflation 0.043 0.004

(9.30)** (20.43)**Aggregate price dispersion across channels -1.328 0.005

(3.03)** (0.22)Corralito premium*credit card ownership 0.009

(0.84)Pseudo-level inflation 0.035

(1.16)Pseudo-level price dispersion 0.477

(0.32)Total household labor hours worked -0.005

(1.54)Constant 6.132 4.379 7.554 0.392

(12.28)** (8.56)** (19.87)** (20.92)**

Pseudo-household fixed effects yes yes yes yesTime effects yes yes no noProduct quantity effects yes yes yes no

How prevalent is this for crisis mitigation? One of the most prevalent mechanisms

66% households use it.60%-80% of households use different

consumption strategies. Instead, 13% use labor market, and between

2 and 11% use formal or informal credit, although adjustments in the labor and credit market have received much more attention in the literature.

TABLE 7: PREVALENCE OF USE OF DIFFERENT ADJUSTMENT MECHANISMS

Adjustment Mechanism AllLowest 2nd 3rd 4th Highest

Shopping Frequency from LatinPanel databaseIncrease in days 61.6 60.3 56.3 61.3 61.6 66.2Increase in channels 75.8 72.6 71.8 80.0 79.5 74.3Increase in channel-days 66.0 61.6 63.4 65.3 65.8 71.6

World Bank Survey on Crisis Coping Strategies:Consumption StrategiesReduced quantity of food 74.9 90.4 83.1 73.2 69.0 59.1Substituted for cheaper food 92.3 97.6 95.4 92.5 91.5 84.8Reduced consumption of non-food items 81.0 90.5 87.7 81.5 76.8 68.3Substituted non-food items for cheaper items 83.2 89.5 89.3 80.4 80.2 76.6Increased home production 61.1 64.4 73.0 62.6 52.5 43.2

Labor Market StrategiesAdding new workers to labor market 12.9 28.0 16.8 12.2 6.2 1.4Working more hours 13.7 11.4 15.6 16.3 11.5 13.4

Financial StrategiesSelling assets 3.3 5.9 3.7 3.3 2.7 1.1Using savings 4.8 2.8 3.5 4.0 8.0 5.6Borrowing from banks 1.8 0.9 3.6 1.8 0.6 2.0Borrowing from friends and family 11.3 21.2 15.7 10.6 5.8 3.0Purchase with delayed payment 8.0 14.6 13.1 9.5 2.3 0.7

Percentage of Households using:Income Quintile

Why shop more?

To search out lower prices for the same products

Instrument channel-days with EPH income

htqihhttqihtqi sChanneldayprice ,,,,,,,,,ln

Results

Shopping one more channel-day results in a 18% fall in the price of same product, along with 2% saving from switching to cheaper brands

Crisis is estimated to increase channel-days by 0.20, so estimated 4% saving in price

Expenditure fell 10%, so shopping increase mitigates approximately 40% of fall in expenditure through cheaper prices.

Conclusions Argentine consumers increased their shopping

frequency as means of coping with the crisis Showed a mitigation mechanism that standard

income and expenditure surveys provide little information about

Mechanism is used by many households and has significant effects

In the presence of aggregate shocks which prevent households from taking their labor to the labor market, non-market uses of labor like searching for better prices provide an alternate way for households to smooth consumption.